Computer vision for waste sorting and recycling contamination reduction: A U.S application

Recycling contamination remains a persistent barrier to efficient material recovery in U.S. material recovery facilities, reducing bale quality, increasing processing costs, and limiting environmental gains. Although computer vision and artificial intelligence systems have been introduced to improve sorting performance, the literature remains fragmented, with limited synthesis of technical capabilities, economic viability, and environmental implications within a unified U.S. context. This study presents a systematic literature review that addresses that gap by evaluating recent advances in vision-based waste classification, robotic sorting integration, contamination detection, and AI-enabled performance optimization. The study critically examines model architecture, dataset limitations, real-time constraints, and reported industrial deployments, while also assessing economic and policy implications. Findings indicate that AI consistently improves contamination reduction and material recovery in secondary sorting lines, yet challenges persist in generalization, dataset standardization, and cost scalability. The paper provides an integrated framework linking technical performance, operational economics, and environmental impact to guide future research and deployment strategies.